Hits:

Indexed by:期刊论文

Date of Publication:2010-11-01

Journal:INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL

Included Journals:SCIE、EI、Scopus

Volume:6

Issue:11

Page Number:5093-5103

ISSN No.:1349-4198

Key Words:Protein subnuclear location; Elman networks ensemble; Amphiphilic PseAA ammo acid composition; Re-substitution test; Jackkinfe test

Abstract:Knowledge of nuclear-protein localizations plays a very important role in understanding the biochemical processes of the nucleus With the avalanche of protein sequences generated in the post-qenomic era, an automated method is badly needed to annotate the subnuelear locations of numerous newly found nuclear protein sequences in a short time In this paper, a novel approach is developed for predicting the protein subnuclear locations Firstly, a powerful ensemble classifier based on the Elman networks is proposed Secondly, the protein samples are represented by amphiphilic pseudo amino acid (PseAA) composition, which can incorporate a considerable amount of sequence-order effects Thirdly, six differ cut algorithms are adopted to consider the diversity of the base ensemble, and 18 Elman networks are subsequently obtained by using three different node numbers of neurons in the hidden layers Lastly, as a demonstration, identifications are performed for 9 subnuclear locations in 714 nuclear proteins The accuracy rates, obtained in both a re-substitution test and a jackknife test, are significantly higher than those achieved by other classifiers It is anticipated that the proposed approach may become a useful tool to reduce the huge gap between the number of gene sequences in databases and the number of gene products which have been functionally characterized